A model that doesn't fit doesn't run slowly, it doesn't run. Here's every card from our bench, sorted by measured speed.
Weights: unsloth/LFM2.5-8B-A1B-GGUF
| GPU | Speed | VRAM | Basis |
|---|---|---|---|
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 617.09 tok/s | 96GB | ✓ Measured |
| NVIDIA B300 | 596.16 tok/s | 288GB | ✓ Measured |
| NVIDIA H200 | 583.96 tok/s | 141GB | ✓ Measured |
| NVIDIA H100 80GB HBM3 | 581.98 tok/s | 80GB | ✓ Measured |
| NVIDIA B200 | 575.15 tok/s | 192GB | ✓ Measured |
| NVIDIA L40S | 394.48 tok/s | 48GB | ✓ Measured |
| NVIDIA A100 80GB SXM4 | 369.01 tok/s | 80GB | ✓ Measured |
| NVIDIA A100 40GB SXM4 | 351.47 tok/s | 40GB | Est. |
| NVIDIA A10G | 268.04 tok/s | 24GB | ✓ Measured |
| NVIDIA L4 | 170.2 tok/s | 24GB | ✓ Measured |
| NVIDIA T4 | 139.03 tok/s | 16GB | ✓ Measured |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition is the fastest card we've measured running LFM2.5-8B-A1B, at 617.09 tok/s.
NVIDIA T4 ($2,299 MSRP) is the cheapest card on our bench that runs LFM2.5-8B-A1B, at 139.03 tok/s.